Create Your First Agent

This walkthrough follows the current dashboard flow in web/app/(dashboard)/agents/create/page.tsx and the follow-up agent workspace in web/app/(dashboard)/agents/[agentName]/view/page.tsx.

Before You Start

Make sure you have:

• the web app running at http://localhost:3005
• a working account that can access the dashboard
• at least one LLM provider key ready

The current create flow will not let you finish without:

• name
• description
• llm_provider
• llm_model
• llm_api_key
• system_prompt

Step 1: Open the Create Flow

1. Sign in to Synkora.
2. Open Agents.
3. Click Create Agent.

The create screen is a four-step wizard:

1. Basics
2. AI Model
3. Personality
4. Review

Step 2: Complete Basics

On the Basics step, fill in the agent identity first:

1. Enter the agent name.
2. Add a short description of the job it should do.
3. Optionally upload an avatar.
4. Optionally choose a template or public-facing settings if they fit your use case.

Keep the first version narrow. Good first agents are usually one of these:

• support assistant
• documentation assistant
• internal operations helper
• research assistant

If the name already exists, the create flow returns you to Basics and shows the conflict there.

Step 3: Configure the AI Model

On the AI Model step:

1. Select the provider.
2. Select the model.
3. Paste the provider API key.
4. Adjust optional fields only if you need them, such as API base, temperature, max tokens, or top_p.

This is the minimum model configuration the current UI enforces before you can continue.

Step 4: Set Personality and Instructions

On the Personality step:

1. Choose one of the built-in presets:
   • Friendly
   • Professional
   • Expert
   • Custom
2. Review the generated system prompt.
3. Replace it with your own prompt if needed.

The wizard requires a non-empty system prompt before it will move forward.

Step 5: Review and Create

On Review, confirm the configuration and create the agent.

The current create flow does two things:

1. Creates the agent record and its initial LLM config.
2. If you selected capabilities, enables them in bulk after the agent is created.

On success, Synkora redirects to:

/agents/{createdAgentSlug}/edit?tab=llm-models

Step 6: Validate the Agent

After creation, use this order instead of configuring every feature at once.

1. Test the core behavior

Open:

/agents/{agentName}/chat

Send a few prompts and confirm the response quality before adding knowledge or tools.

2. Review the model config

Open:

/agents/{agentName}/llm-configs

Use this page when you need to change provider, model, or model settings after the initial setup.

3. Attach knowledge

Open:

/agents/{agentName}/knowledge-bases

Use this when the agent needs grounded answers from your documents or site content.

See Create a RAG Agent.

4. Add tools

Open:

/agents/{agentName}/tools

This is where you turn the agent from “answers questions” into “takes action”.

See Add Tools.

5. Pick delivery surfaces

The current agent workspace exposes these deployment and delivery paths:

• landing-page
• widgets
• domains
• api-keys
• slack-bots
• messaging-bots
• telegram-bots
• webhooks
• voice

Pick one surface first. Do not configure everything on day one.

6. Turn on advanced features only when needed

The current UI also exposes:

• autonomous
• sub-agents
• mcp-servers
• outputs
• chat-customization
• lens
• database-connections
• compute
• subscriptions

These are useful, but they should come after the base agent behavior is stable.

Recommended First-Agent Sequence

For most teams, the cleanest path is:

1. Create the agent.
2. Test it in chat.
3. Attach one knowledge base if it needs context.
4. Add one or two tools.
5. Expose it through a widget, API key, or one messaging channel.
6. Watch usage and quality in Lens.

Related Pages

• Agents
• Knowledge Bases
• Embed the Web Widget